Arc Flash Hazard Analysis
for Data Centres:
IEEE 1584-2018 Methodology,
Incident Energy & PPE Selection

In 2018, IEEE published the most significant revision to arc flash calculation methodology in the standard’s history. IEEE 1584-2018 replaced the empirical equations of the 2002 edition — which were based on limited test data and known to produce inaccurate results for many common equipment configurations — with a new model derived from over 1,800 arc flash tests across a wide range of equipment types, bus gaps, and enclosure geometries.

The consequence for data centre operators and electrical engineers is direct: arc flash incident energy calculations performed using the 2002 standard may be significantly incorrect — either dangerously underestimating the arc energy at some equipment or unnecessarily overestimating it at others. Any data centre that has not had an arc flash study performed or reviewed using the 2018 standard is operating with potentially incorrect PPE specifications and protection boundaries.

This article explains the arc flash phenomenon, the IEEE 1584-2018 methodology, the data centre-specific parameters that drive arc flash severity, the PPE selection process, and the engineering mitigation strategies that reduce arc flash hazard at the design stage — before anyone is required to work on energised equipment.

What Is Arc Flash and Why Does It Matter for Data Centres

An arc flash occurs when electrical current travels through ionised air between energised conductors or between a conductor and earth. The arc plasma reaches temperatures of 19,000–35,000°C — hot enough to vapourise copper conductors instantly, producing an explosive pressure wave and a radiant heat release that causes severe burns to anyone within the arc flash boundary, regardless of whether they are in direct contact with the electrical equipment.

Data centres have specific arc flash characteristics that make them a higher hazard than many industrial environments:

• High fault levels at LV busbars — parallel transformer operation can produce 50–130 kA at 415 V LV busbars, as shown in the companion SCCAF article. Higher fault current means more arc energy per unit time.
• Frequent live electrical work — meter reading, panel inspection, PDU connection, UPS maintenance, generator testing — data centres require more frequent access to energised electrical equipment than most industrial facilities, increasing exposure probability.
• 24/7 operations constraint — the cost of de-energising a live data centre for electrical maintenance is enormous, creating commercial pressure to perform more work on energised equipment than would be acceptable in facilities that can tolerate scheduled downtime.
• High-density power electronics — UPS systems, PDUs, and battery systems all present arc flash hazards that differ from conventional switchgear in ways the 2002 IEEE standard did not adequately capture.

IEEE 1584-2018 Methodology

IEEE 1584-2018 uses a multi-variable empirical model based on laboratory test data to calculate arc flash incident energy. The key inputs are the arcing fault current, the arc duration (determined by protection clearing time), the working distance, the equipment type (open bus, switchgear, MCC, cable junction box), the bus gap, and the enclosure dimensions.

Key Equations and Parameters

// IEEE 1584-2018 — Arc Flash Calculation Overview
// (Simplified — full model uses multi-variable regression equations)

// Step 1: Calculate arcing fault current Iarc
// Iarc is less than bolted fault current Ibf due to arc impedance
// IEEE 1584-2018 provides empirical equation based on Ibf, voltage, and gap
Iarc ≈ 0.6 × Ibf  to  0.85 × Ibf  (typical range for LV systems — use software for exact value)

// Step 2: Determine arc duration t from protection time-current study
// This is the TIME the arc burns before the protecting device clears it
// t = relay operating time + breaker clearing time
// Example: IDMT relay (0.3 s) + ACB (0.08 s) = 0.38 s arc duration

// Step 3: Calculate incident energy E at working distance D
E = 4.184 × Cf × En × ( t / 0.2 ) × ( 610^x / D^x )   (IEEE 1584-2002 simplified form)

// IEEE 1584-2018 replaces this with enclosure-specific regression model:
// E = f( Ibf, V, gap, enclosure type, enclosure dimensions, D, t )
// Cannot be manually calculated — requires software implementation of the 2018 model

// Step 4: Convert to PPE category
E (cal/cm²)  → PPE category (NFPA 70E Table 130.5(G))
0–1.2       → Category 1  (4 cal/cm² minimum arc flash PPE)
1.2–12      → Category 2  (8 cal/cm² minimum arc flash PPE)
12–40       → Category 3  (25 cal/cm² minimum arc flash PPE)
40–100      → Category 4  (40 cal/cm² minimum arc flash PPE)
>100        → Danger — no PPE rated for this exposure; de-energise before working

Arc Duration: The Most Critical Variable

Incident energy is directly proportional to arc duration — doubling the arc duration doubles the incident energy at the worker’s position. Arc duration is determined entirely by the protection system: how quickly does the upstream protective device detect the arcing fault current and interrupt it?

This is why arc flash analysis and protection coordination are inseparable engineering activities. Every protection relay time setting, every circuit breaker trip curve, and every protection coordination decision directly affects the arc flash incident energy at every panel and switchboard in the facility.

Worked Example: Data Centre Main LV Panel

// Arc Flash worked example — 5 MW data centre main LV panel
// From SCCAF: I"k = 108 kA at LV busbar (two transformers in parallel)
// Protection: IDMT relay, TMS = 0.1, pick-up = 2×In, at 10×In → t = 0.35 s
// ACB clearing time: 0.08 s  |  Total arc duration: 0.43 s
// Working distance: 610 mm (typical LV panel per IEEE 1584)
// Equipment type: LV switchgear (metal-enclosed)
// Bus gap: 32 mm (standard 415 V LV switchgear)

// Step 1: Arcing fault current (IEEE 1584-2018 — software calculated)
Iarc   = ~72 kA  (at 415 V, 32 mm gap — approximately 0.67 × I"k)

// Step 2: Incident energy (IEEE 1584-2018 software output)
E      = ~85 cal/cm²  at 610 mm working distance

// Step 3: PPE category assignment
// E = 85 cal/cm² → EXCEEDS Category 4 maximum (40 cal/cm²)
// Result: DANGER — no PPE rated; must de-energise before working

// ─────────────────────────────────────────────────────────────
// MITIGATION OPTION 1: Add arc flash detection relay
// New arc duration: 10 ms (AFD) + 40 ms (breaker) = 50 ms
E_AFD  = ~10 cal/cm²  → Category 2  (8 cal/cm² PPE — manageable)

// MITIGATION OPTION 2: Operate in split-bus (one transformer only)
// New I"k: 55 kA (single transformer)  |  Iarc ≈ 37 kA
E_split= ~22 cal/cm²  → Category 3  (25 cal/cm² PPE — heavy but manageable)

// MITIGATION OPTION 3: Higher impedance transformers (%Z = 8%)
// New I"k: 68 kA (parallel)  |  Iarc ≈ 46 kA
E_highZ= ~55 cal/cm²  → Still DANGER without additional mitigation

// RECOMMENDATION: AFD relay is the most effective single mitigation at this panel

PPE Selection: Categories, Materials, and Data Centre Practice

PPE selection for electrical work is governed by NFPA 70E (Standard for Electrical Safety in the Workplace) in the USA — and while India does not have an equivalent national standard that is as detailed, NFPA 70E is widely adopted as the engineering reference by multinational data centre operators in India. IS 5216 and IS 4770 provide some guidance but do not include arc flash-specific PPE categorisation.

PPE Limitations That Data Centre Operators Must Understand

Arc flash PPE is rated for a specific incident energy level. It provides a 50% probability of a second-degree burn (curable) at its rated incident energy — not zero burn probability. At incident energies exceeding the PPE rating, PPE failure and fatal injury are likely. This means that PPE is a last-resort protection measure — it is not a substitute for reducing arc flash incident energy through engineering controls (fast protection, arc flash detection relays, remote racking, infrared windows).

Engineering Mitigation: Reducing Incident Energy by Design

The most effective arc flash mitigation strategies are engineering controls that reduce incident energy at the source — rather than relying on workers wearing heavy PPE. These should be evaluated and specified during the data centre design phase, not added reactively after an arc flash study reveals high incident energies.

• 01

  Arc Flash Detection Relays (AFD)

  As shown in the worked example, AFD relays reduce arc duration to 10–15 ms by detecting the arc light directly rather than waiting for overcurrent relay operating time. The incident energy reduction is 20–40× compared to IDMT relay protection — transforming a Category 4 or Danger situation into a Category 1 or 2. AFD relays should be standard specification for all main LV switchboards and 33 kV indoor switchrooms in data centres. Cost: approximately ₹1.5–3 L per switchboard — negligible against the capital cost of the switchgear it protects.

• 02

  High-Impedance Transformers

  Increasing transformer %Z from 5–6% to 7–8% reduces the available fault current and therefore the arc flash incident energy. The trade-off is a larger voltage drop under full load and slightly higher transformer cost. For data centres where LV incident energy is very high due to parallel transformers, specifying %Z = 8% can reduce incident energy by 20–30% — worthwhile when combined with other measures but rarely sufficient as a sole mitigation.

• 03

  Bus Differential Protection (87B)

  Differential protection on the LV main busbar uses current transformers on each incomer and the bus coupler to detect busbar faults within milliseconds — without waiting for overcurrent relay time delays. Bus differential (87B) is the most effective overcurrent-based protection for main LV busbars where multiple incomers make graded overcurrent protection difficult to optimise. Reduces arc duration to 40–60 ms (breaker clearing time only) without requiring arc light sensors.

• 04

  Infrared (IR) Windows on Switchgear

  IR windows installed in switchgear panels allow thermographic scanning of busbar connections and cable terminations without opening the panel. Thermal anomalies (loose connections, overloaded cables) are the most common precursors to arc flash events — detecting them with quarterly IR scanning eliminates a significant proportion of arc flash initiating causes. IR windows are a low-cost addition at panel manufacturing stage (₹3,000–8,000 per panel) that is very expensive to retrofit.

• 05

  Remote Racking / Remote Operated Switches

  Withdrawable circuit breaker racking (moving breakers between service and test positions) has historically required the operator to be in front of an open panel — the highest-risk electrical activity in a data centre. Remote racking devices allow this operation to be performed from outside the arc flash boundary. For data centres specifying new 33 kV GIS or MV switchgear, specify motorised switching and remote operation as standard — eliminating the need for anyone to be present during switching operations.

• 06

  Maintenance Mode on Protection Relays

  Temporarily reducing protection relay time delays — “maintenance mode” on digital relays — during planned electrical work reduces arc duration without requiring permanent changes to the coordination scheme. During the maintenance window, the upstream relay is set to instantaneous trip; after work is complete, normal graded settings are restored. Some digital relay platforms automate maintenance mode activation — triggered by a key switch or authorised command — and provide an automatic time-out to prevent operators from forgetting to restore normal settings.

Arc Flash Labels and Safety Documentation

Every electrical panel, switchboard, and distribution board in a data centre must be labelled with the results of the arc flash study — the incident energy, PPE category, working distance, and restricted approach boundary. These labels are not decorative compliance items; they are the communication mechanism that tells every person working near that equipment what hazard they face and what protection they need.

Data Centre-Specific Arc Flash Considerations

UPS Systems and Inverter-Fed Panels

UPS output panels present a unique arc flash calculation challenge. As discussed in the SCCAF companion article, the UPS limits fault current to 1.0–1.5× rated for a defined time before transferring to bypass. During normal (inverter) operation, the arc flash incident energy at the UPS output panel may be very low — limited by the inverter current capability. But when the UPS transfers to static bypass — which happens automatically during maintenance or overload — the full upstream fault level is presented to the output panel instantly. The arc flash label at UPS output panels must reflect the bypass mode incident energy, not just the inverter mode energy, since bypass operation is the condition under which maintenance is typically performed.

33 kV / 11 kV Switchgear Rooms

HV switchgear rooms in data centres — 33 kV GIS substations and 11 kV MV distribution switchboards — present very high arc flash hazard at relatively low fault currents due to the high voltage. The incident energy at 11 kV with modest fault current and a 0.5 s protection clearing time can easily exceed 100 cal/cm² — well into the “Danger” zone. For HV switchgear rooms in data centres, the engineering response is:

• Arc flash detection relays as standard — not optional
• All normal switching operations performed remotely (motorised switches)
• No live work inside energised HV panels without demonstrated incident energy below Category 2 through combination of AFD and fast clearing time
• Regular SF₆ gas density monitoring on GIS — SF₆ loss is an early indicator of potential internal arc development

Battery Rooms — UPS VRLA and BESS

Battery rooms present DC arc flash hazard in addition to AC arc flash. DC arc flash differs critically from AC arc flash: AC arcs naturally extinguish at each zero-crossing (100 times per second at 50 Hz); DC arcs are sustained and do not self-extinguish. A DC arc at 400 V bus from a 10 MW BESS can sustain at currents where an equivalent AC arc would self-extinguish, resulting in longer arc duration and higher incident energy. IEEE 1584-2018 covers AC arc flash; DC arc flash requires separate analysis per IEEE 1584 DC supplement or NFPA 70E DC guidance. Ensure the arc flash study scope explicitly includes all DC bus connections in the BESS room and UPS battery strings.

Study Scope, Triggers, and Review Cycle

Conclusion: Arc Flash Analysis Is a Personnel Safety Obligation

Arc flash analysis using IEEE 1584-2018 is not a regulatory box-ticking exercise — it is the engineering foundation for protecting the people who maintain the electrical infrastructure that keeps data centres running. Those people are at risk every time they open a panel door, read a meter, or perform switching operations on energised equipment. Without a current arc flash study, they are working in the dark — literally unaware of the energy that could be released in the event of an arc initiation.

The engineering mitigation strategies available — arc flash detection relays, bus differential protection, remote operation, maintenance mode protection settings — can reduce incident energy dramatically at relatively modest cost when specified at design stage. Their cost increases significantly when retrofitted to operating facilities. The correct time to design for arc flash safety is before construction begins, not after an incident makes it urgent.

An arc flash study that drives the right engineering decisions at design stage is the most cost-effective safety investment a data centre can make. The alternative — PPE, training, and emergency response — is more expensive, less reliable, and provides far less protection for the people doing the work.